In short
- Scene photos are generated by third-party AI models from your photo and a scene recipe.
- Crops, white-background images and the standard motion clips are made from your own photo without AI generation.
- You review and approve every result. AI can change product details.
Where we use AI
| Feature | What the AI does | Provider |
|---|---|---|
| Scene photos | Generates a new image of your product in the scene you choose, from your photo and the scene recipe | Image models via OpenRouter or fal.ai |
| Product facts | Reads your photo to suggest details worth preserving, which you can edit | Vision model via OpenRouter |
| Selling packs | Only the frames marked as quoted scene frames are generated; crops and white-background images are made locally from your own photo | Same as scene photos |
| Motion | The name overlay and before/after clips are assembled locally from a still you approved. An optional experimental AI camera move, when enabled, is generated from that still and may alter details | None for local clips; fal.ai for the experimental preset |
How AI content is labelled
In the studio, every generated frame is labelled as a generated variant and shown next to your source photo. Generations from untested scene recipes carry an extra warning. On this website, sample images are marked as AI concept studies, and cards without an image are drawn set studies, not photographs.
Downloaded files do not yet carry a machine-readable marker that identifies them as AI-generated. This is a known gap. Until it is closed, label AI-generated images yourself wherever the law or a platform requires it.
Known limits
- Models can change small text, logos, proportions, colours and materials, even when told to preserve them.
- A scene can contain objects or lighting that would be impossible in a real photo.
- Results are not reviewed by us before you see them. You decide which frames to keep through the approve and reject controls.
Human control
Nothing is published or shared automatically. You choose the source photo, the scene and how strictly the product is preserved, and you review every result before downloading or adding it to a pack. A saved cutout lets you mark exactly which pixels belong to the product.
Your data and model training
We do not train our own models on your photos or results. Photos are sent to a provider only when you start a request. How each provider handles that data is described on the service providers page and in the privacy policy.
When you publish results
Depending on where and how you publish, you may have to tell viewers that an image was generated or edited with AI, for example under the EU AI Act or a marketplace’s rules. Our acceptable use policy explains how to keep product images honest.